Real People.
Real Money. Gone.

The confusion is by design — most people don't know that a UPI PIN only ever authorises payment out, never in. By the time the app's own confirmation screen appears, the transaction is one tap from irreversible.

This is what a predictive pre-transaction gate is for — scoring device and session risk before the transaction executes, flagging a high-risk pattern (a first-time payee, an unusual context) before PIN entry rather than after. The resulting hash-chained evidence record also settles the "what did I actually authorise" dispute after the fact.

The victim is a genuine customer, on their own enrolled device, moving their own money — willingly, if under manufactured panic. No cryptographic check is designed to catch this, because nothing about the transaction's mechanics is wrong.

The behavioural layer has a named category for exactly this — Coercion-Specific detection for digital-arrest and romance-scam-style pressure — watching for the session-level signs of duress rather than trusting that a correctly-signed transaction means a willing one. It's the same honest admission this platform makes everywhere else: possession proof can't stop this; surgical behavioural friction has to.

A VPN makes a connection from Cambodia look like it's coming from anywhere the fraudster chooses. Systems that trust an IP address for location or reputation checks have nothing left to check once that address is rented, not real.

This is what network and edge analysis is for — ASN reputation blocking and VPN/proxy detection at the edge, plus GPS-vs-IP cross-checks that catch a device claiming to be somewhere its actual network path contradicts. The IP alone was never the trust signal; where it disagrees with everything else is.

A sideloaded app, installed outside the Play Store, with permissions to read notifications and SMS — nothing about that requires rooting the phone, so a check that only looks for root or jailbreak state misses it entirely.

RASP's Tamper, Repackaging & SDK Self-Integrity and Hooking Framework detection categories catch this class of compromise directly, and the device fingerprint's own installer-source field flags that the app arrived from outside the Play Store in the first place — a fact the OS knows and most banking apps never ask it.

A remote-access session isn't malware in the traditional sense — the victim installed it willingly. Most banking apps have no way to tell "I'm being viewed and controlled by someone else right now" from a normal session.

This is a named detection category in the RASP layer — Screen-Share, Overlay & UI Attacks — checked as part of the 100+ on-device vectors every session. An active screen-mirroring or remote-control session is a detectable device state, independent of whether the app installed was "legitimate" software like a remote-access tool.

An OTP sent by SMS is only as trustworthy as the SIM it lands on. Once the number itself is stolen, the OTP arrives at the fraudster's phone — a perfectly valid code, on the wrong device entirely.

This is precisely what hardware-bound identity is for. The signing key that authorises a transaction lives inside the original device's own security chip — it does not move with a phone number to a new SIM or a new phone. A SIM swap gets the fraudster the OTP channel; it does not get them the key. No key, no valid signature, no transaction.

A name, a PAN, and a photo are all easy to fabricate or steal. Underwriting systems built to check identity documents have no way to notice that the phone submitting the tenth application this week is the same phone that submitted the first nine — under nine different names.

Device fingerprinting exists for exactly this: a stable, hardware-anchored device identity that survives app reinstalls and cleared data, so a lender can catch one device applying under many names before disbursal — not after the money is already gone and the harassment has already begun.

The app's interface was the entire con — a real-looking balance, a real-looking chart, no real exchange underneath any of it. Nothing about the victim's device was compromised; the fake was the product itself.

For a genuine broker protecting its real customers, this cuts the other way: APK certificate fingerprinting and tamper detection confirm a customer is running the broker's actual signed app, not a repackaged clone — and the same device identity that catches loan-stacking also flags a device pattern seen across other reported scam apps.

Every individual transfer looked like ordinary behaviour from a trusted device — the victim's own phone, sending money the victim chose to send. The tell isn't any single transaction; it's the escalating pattern across all of them.

This is what a per-device behavioural baseline is built to catch — a session-and-transaction pattern (rapid escalation, an entirely new payee category, timing that doesn't match the device's own history) flagged as anomalous even though every individual step looks clean.

Urgency plus an official-looking bank name is usually enough. The link itself often leads to a sideloaded app or a credential-harvesting page dressed up as the bank's own KYC portal.

The same fingerprinting fields that catch banking trojans apply here — installer source and APK certificate checks flag anything sideloaded outside the Play Store as exactly that, before it gets anywhere near a real session.